1

Cooperative Step-wise Relaying and

Combining for Multi-hop Vehicular Wireless
Communication
Yi Zhu, Student Member, IEEE, Lixing Song, Student Member, IEEE, Shaoen Wu, Member, IEEE
Honggang Wang, Senior Member, IEEE and Chonggang Wang, Senior Member, IEEE
Technical Report V3

Abstract
Vehicular ad hoc networks (VANETs) require cooperation among vehicles in exchanging information. Cooperative
communication among these vehicles can significantly improve wireless communication performance. This paper
proposes a novel cooperative scheme named Cooperative Step-wise Relaying and Combining (CSRC) for multihop wireless networks, which utilizes low-order demodulations to patricianly demodulate high-order modulated
signal cooperatively at intermediate relay nodes. The bit error rate probability is analyzed. The comparison among
the proposed scheme and the conventional scheme is conducted through simulation and software defined radio
experiments. Results show that the proposed scheme has great potential in the performance improvement of cooperative
communication.
Index Terms
VANET, Cooperative Communication, Multi-hop wireless networking.

I. I NTRODUCTION
VANETs depend on wireless medium for communication. Wireless communication suffers signal variation and
degradation from various causes such as multi-path fading, path loss and mobility. The signal variation and
degradation can be mitigated by exploiting user diversity unique in wireless networking where users may experience
different link conditions because of their locations and speeds [1]. In recent years, novel opportunistic technologies
of cooperation among wireless mobile terminals, referred as cooperative relay, have been considered as a potential
Yi Zhu is with the School of Computing, University of Southern Mississippi, Hattiesburg, MS, 39406 USA. E-mail: zhuyi890108@gmail.com.
Lixing Song and Shaoen Wu are with the Department of Computer Science, Ball State University, Muncie, IN 47306 USA. E-mails: {slong,
swu}@bsu.edu.
Honggang Wang is with Department of Electrical and Computer Engineering, University of Massachusetts Dartmouth, Dartmouth, MA 02747
USA. E-mail: hwang1@umassd.edu.
Chonggang Wang is with InterDigital Communications, Wilmington, DE 19809 USA.

November 25, 2013

DRAFT

scheme to significantly improve the communication system performance by cooperatively relaying the received
signals at intermediate nodes to the destination in order to constitute a distributed virtual multiple input multiple
output (MIMO) system [2][4]. Because of various gains of cooperative relay over the conventional single-path
transmission, it is expected to be an essential technology to the next generation wireless networks. In particular,
relay station in cellular system is able to increase energy efficiency and extend cell coverage [5]. As for VANETs,
because it has been shown that vehicles on road tend to form clusters in driving, a cluster of connected vehicles
offers the possibility for them to cooperate in communication.
In wireless communication, normally, high-order modulations lead to large bit error rate (BER) in wireless
communication, especially under low SNR channel conditions. Although low-order modulations can address the
problem, they yield low bit rate. The motivation of this work is to design a cooperative approach that is capable of
maintaining the high bit rate of high-order modulations while keeping the low error rate of low-order modulations.
In this paper, we propose a novel cooperative strategy called Cooperative Step-wise Relaying and Combining
(CSRC) for multi-hop wireless networking system. CSRC partially demodulates the signal symbol with low-order
robust modulations, let us say QPSK, at each cooperative relay node. A destination node combines all the partial
demodulated outcomes to correctly decode the symbol. Unlike conventional cooperative communication strategies
that are generally based on amplify-and-forward or decode-and-forward relaying schemes, CSRC makes a partial
demodulation decision at each relay node, and relays the signal and decision to the successive node(s). In particular,
this strategy supports high-order modulations, such as 16-QAM or 64-QAM, for high bit rates over multi-hop
wireless networks.
The bit error probability of CSRC is theoretically analyzed. One important observation is that the collaborative
gain can reach its capacity at a certain channel condition. Extensive simulation evaluations are conducted and prove
the strengths of the proposed CSRC. Furthermore, the comparisons of performance between the proposed CSRC
and conventional schemes are conducted on a software defined radio (SDR) testbed. The results of both simulation
and SDR testbed experiments show the great potentials of CSRC in improving the communication performance.
In the rest of this paper, the related cooperative relay work is reviewed in Section II and the system is modeled
in Section III. Then, Section IV presents the proposed Cooperative Step-Wise Relaying and Combining . In
Section V, the bit error probability of CSRC is analyzed. The performance evaluation on simulation and SDR
testbed experiments are presented in Section VI. Finally, Section VII concludes this work.
II. R ELATED W ORK
There are two fundamental components in cooperative communication: relaying and combining. The typical
approach for cooperative communication is to relay the received signal at intermediate nodes and finally combine
the multiple received copies of the signal at the destination node. Different in how the information is processed in
the relaying, various schemes have been proposed in the literature, including amplify-and-forward [6], demodulateand-forward [7], decode-and-forward [6], [8] and soft decode-and-forward [9]. In amplify-and-forward (AF), a relay
node amplifies the received signals and forwards the scaled signal to the destination. In demodulate-and-forward,
November 25, 2013

DRAFT

a relay node demodulates the received signals and forwards regenerated signals to the destination. In decode-andforward (DF), a relay node decodes the received signals, re-encodes and forwards the regenerated signals to the
destination. It has been remarked that AF has low-complexity and significantly saves transmission power. However,
compared to DF, AF has two main drawbacks. One is that it does not have coding gains and the other is that it also
amplifies and forwards noises. DF has the advantages of regenerating the signal, and correcting errors at the relay.
Nevertheless, when the capability of error correcting in the decoding is not strong enough to correct all errors, the
errors will be propagated throughout the network.
Azarian, Gamal and Schniter investigated the diversity-multiplexing tradeoff of relay protocols [10]. Coding
schemes such as distributed Turbo codes [11], [12] have been also studied to exploit the cooperative diversity for
DF in relay channels. Recently, Low Density Parity Check (LDPC) codes were investigated for half-duplex relay
channels [13], [14]. In the case that the channel is not reliable enough to guarantee error-free decoded bits at the relay
nodes, avoiding error propagation in the relaying becomes challenging. To address this challenge, relay schemes
that attempt to combine the benefit of DF and AF were proposed, such as the soft decode-and-forward protocols [9],
[11]. In soft decode-and-forward, the soft information in decoding the source signal is used to form a soft signal at
the relay based on the log-likelihood ratios. Then, in addition to the raw signal, the soft signal is transmitted to the
destination node with relaying schemes of AF, such as Estimate-and-Forward [15] and Decode-Estimate-Forward
[16] , which use the minimum mean square error estimation (MMSE) to optimize SNR.
In cooperative communication, the source and relay signals are finally combined at the destination. A possible
combining technique is the well-known Maximum-ratio-combining (MRC). However, it is suboptimal and does not
achieve full diversity [17]. Serious performance degradation of MRC can be caused by the error propagation at the
relays. Another important challenge is that MRC incurs high computational complexity at the destination, especially
when high-order modulations are employed. Wang, et al. proposed Cooperative-Maximum Ratio Combining (CMRC) to achieve full diversity with DF relaying schemes by exploiting the knowledge of the instantaneous bit-error
probability of the source-relay link at the destination [18].
All of the relay protocols that have been discussed above suffer from the problem of low spectrum efficiency.
Specifically, their spectrum efficiency is half of that of direct transmission. To improve the spectrum efficiency,
two-way relaying is proposed [19], [20]. The packets from two ends are combined at the relay node. The combined
packets are relayed to two ends. Finally, these two can extracts the packets they need by subtracting the packets
they have sent. In three terminal and TDMA setting, two-way relaying need two slots instead of four for one-way
relay, which is similar to network coding [21].
III. S YSTEM M ODEL
This section describes the system model and assumptions that this paper is based on. Consider the half-duplex
multi-hop relay system shown in Figure 1, there are two topologies. The upper one is a one-branch relay network
model; the lower one is an N -branch cooperative network model. The one-branch model is used to facilitate the
explanation of the N -branch system. Both of them consist of one source, one destination, and distributed relays
November 25, 2013

DRAFT

that cooperatively support the communication between the source S and destination D. In this system, the source
encodes the information bits and transmits the modulated signals to the relay node(s), which demodulates, decodes
and re-encodes the received message. The resulting message is then modulated and forwarded to its successive
prospective relay node or the destination node. Finally, the destination receives the signals from the last relay
nodes. For the N -branch collaborative communication model, Maximal Ratio Combining (MRC) is assumed as the
receive diversity at the destination. It is assumed that every node in this system can only hear the two nodes next
to it. In addition, this relay system is an inband setting, namely, transmitting and receiving over the same frequency
band in a half-duplex mode, which is achieved with a time division duplex (TDD), where data are transmitted and
received in separate time slots. We suppose the channels are quasi-static and independent over time slots and all
receiving nodes have instantaneous channel state information (CSI). In order to avoid the interference between links,
time division multiple access (TDMA) is used for providing orthogonal channels to facilitate the communications
between relays.

R1
R2
S

Rn
Fig. 1.

Topologies of channel model

At the source node S, information bits W are encoded and modulated into the signals X {S1 , S2 , , Sk },
 
where Si denotes a symbol, with power constraint, E X 2  P0 . In the baseband model, for all i {1, 2, , n}
corresponding to relays and the destination D, the signal experiences Rayleigh frequency-flat fading characterized
by fading coefficients hi , which are constant complex scalars known to the relays. Zi N (0, i ) captures the
additive white Gaussian noise (AWGN) perceived by nodes, where N (0, i ) denotes symmetric complex Gaussian
distribution with zero mean and variance i . Pi denotes the transmit power and gi denotes the path loss. The

overall channel gain is modeled as Gi = Pi gi hi . The signals forwarded by relays are denoted by Xi for all
i {1, 2, , n}. The received signals at relays and the destination D can be written as:
Yi = Gi Xi + Zi
IV. C OOPERATIVE S TEP - WISE R ELAYING AND C OMBINING
In this section, we explain the proposed Cooperative Step-wise Relaying and Combining (CSRC) that works on
multi-hop networks. In wireless communication, high bit rates can be achieved with high order modulations, such

November 25, 2013

DRAFT

as 16-QAM or 64-QAM, but could result in performance degradation in the case of low channel SNR. To maintain
the high bit rates across poor SNR links, CSRC makes detection with low-order modulations, e.g. QPSK, at each
relay node, and forwards the detection (soft information) to the next node. The detection results are gathered at the
destination to make the final demodulation and decode the information bits. To facilitate the explanation, CSRC with
16-QAM modulation in the one-branch relay network model is elaborated as an example to show how it works.
Then, CSRC is extended for the N -branch collaborative model. In addition, we also discuss the adoption of CSRC
to other high order modulations such as 64-QAM.
A. 16-QAM CSRC in One-Branch Network

Fig. 2.

16-QAM CSRC with three steps

The 16-QAM CSRC for a one-branch network is shown in Figure 2. There are three steps involved. In the first
step, assume that the original bit stream at the source S, is modulated with 16-QAM of square constellation, where
the information is encoded in both amplitude and phase of the transmitted signal. The transmitted signal of 16-QAM
is given by
Si (t) = Ai cos(bi ) cos(2fc t) Ai sin(bi ) sin(2fc t)
,where i {1, 2, 16}. The complex lowpass representation of Si is Si (t) = Re {ui (t)}, where Ui (t) =
SI (t) + jSQ (t) is the equivalent lowpass signal of Si . To facilitate the explanation, without loss of generality, we
assume the transmitted symbol Sr represents 0000, the constellation point at upper left corner on the 16-QAMs
constellation map as in Figure 3 in the first step.

Fig. 3.

The first step: Symbol 0000 is transmitted from the source, S, to the relay node, R.

The the second step is shown in Figure 4. The relay node, R, receives the signal Sr transmitted by the source S.
Unlike any conventional cooperative communication strategy (for instance, amplify-and-forward [6], demodulateand-forward [7], and decode-and-forward [6], [8]), CSRC does not attempt to demodulate the received 16-QAM
November 25, 2013

DRAFT

signal. Rather, it treats the signal as a QPSK modulated signal, which is the core of CSRC. R demodulates the
16-QAM symbol with QPSK. As is shown in Figure 4, at R, the perceived 16-QAM symbol can be detected in
the upper left of constellation map via QPSK demodulation. And then, the relay node R sends two signals to the
destination D. One is the original 16-QAM signal Sr . The other one is the QPSK detection result, a hint, H, which
indicates the location of the original symbol in the constellation map. In this example, the hint indicates the symbol
is in the upper left.
S

Fig. 4.

The second step: S are received by R. S can locate the symbol on the upper left quadrant.

The the third step is shown in Figure 5. In this step, the destination D receives the two signals, Sr and H, from
the relay node R. Before the original symbol Sr is demodulated, H is demodulated first so that Sr is located in
one of the four quadrants of the constellation map. Base on this hint, the 16-QAM modulated Sr can be detected
by QPSK, (NOT 16-QAM), which is more reliable than the detection of 16-QAM. By this moment, the original
signal Sr is fully demodulated with two consecutive QPSK demodulations.

Fig. 5.

The original signal S is demodulated by QPSK incorperating with the hint H.

B. 16-QAM CSRC for Cooperative Networks

Based on the one-branch CSRC, it is easy to extend CSRC to N -branch collaborative multi-hop networks with
treating each branch as a one-brach model. Referring to the N -branch model as in Figure 1, in each individual
branch, the transmitted identical signal is processed with CSRC independently as in the one-branch model. At the
destination, the multiple copies of the raw signals and the hints, Hi , from N branches are combined with any
existing collaborative combing scheme, such as MRC. The CSRC in collaborative case with 16-QAM modulation
is illustrated in Figure 6. The details are as follows, In the first step, the source broadcasts a symbol to N first hop
relay nodes. For simplicity, the symbol of 0000 is used to represent the transmitted signal from the source. In
the second step, all of the first hop relay nodes partially demodulate the symbol, specifically, detecting which of
four quarters where the symbol is located with QPSK. In the third step, relay nodes transmit two signals, which
are the original signal and the partial detection result hint, H. At the destination, with MRC, N original signals

November 25, 2013

DRAFT

are combined, and N partial detection hints are also combined. With the combined partial detections as the final
hint, H, indicating the location of the symbol, the destination fully demodulates the original symbol.

Fig. 6.

16-QAM CSRC for Cooperative Networks

C. Discussion
In addition to16-QAM modulation, CSRC can also be used for other high-order modulations such as 64-QAM
and 256-QAM symbols. The entire procedure for CSRC to work with 64-QAM is illustrated in Figure 7. Based on
the same one-branch three-node model (one source, one relay and one destination), data stream can be modulated
by 64-QAM and transmitted by the source. Then, the relay demodulates the received signal with QPSK to locate the
symbol in one of the four quadrants of the constellation map. This location is modulated by QPSK and forwarded
to the destination with the original 64-QAM signals. At the destination, the original 64-QAM signal is demodulated
by 16-QAM in concert with the QPSK location hint. For CSRC to work with 256-QAM, the relay demodulates
the signal by 64-QAM. The resulted, as a hint, is forwarded with the original 256-QAM signal. The destination
demodulates the original signal by 64-QAM incorporating the 64-QAM hint. Depending on the number of hops
and the modulation scheme, various CSRC procedures are possible in multi-hop networks.

Fig. 7.

CSRC for 64-QAM

CSRC can also be explained from the perspective of the code mapping on the constellation. Consider symbols
are modulated with 16-QAM of square constellation as shown in the left part of Figure 8, which is customized
November 25, 2013

DRAFT

for CSRC. In this code mapping, the first bit indicates that the symbol is on the upper or lower of the plane; the
second bit indicates that a symbol is on the left or the right of the plane; the third bit indicates that a symbol is
on the upper or lower of a 1/4 quadrant; the fourth bit indicates that the symbol is on the left or right of a 1/4
quadrant. As plotted in the right of Figure 8, in the second step of CSRC, the first and second bits are determined
by the replay node R. In the third step, the third and fourth bits are determined by the destination, D. It is clear
that each relay node determines only two bit. Namely, each relay node conducts a QPSK demodulation, which is
much more tolerant to channel noise and fading than demodulating a 16-QAM symbol.

Fig. 8.

The code mapping on constellation and the overview of CSRC for 16-QAM

V. P ERFORMANCE A NALYSIS OF CSRC

This section analyzes the bit error probability of CSRC for 16-QAM with AWGN relay channel to show the
improvement of performance. In the analysis, the channel condition is assumed symmetric across all links. We
consider both one-branch and N -branch collaborative models. In addition, we also derive the performance variation
of CSRC upon the degree of diversity in collaborative communication.
Assume that all channels have the same SNR, 0 . The bit error probability of a symbol with coherent detection
and perfect recovery of the carrier frequency and phase is
p 
20
P0 = Q

(1)

and the symbol error probability of a QPSK is thus

Ps = 1 (1 Q

p 
20 )2

(2)

,where Q (x) is defined as the probability that a Gaussian random variable X with mean 0 and variance 1 is
R
2
greater than x. Q (x) = P (X > x) = x 12 ex /2 dx. AWGN channel can be represented by Binary Symmetric
Channel [22]. The channel transition matrix is

1 P0
T =
P0

P0
1 P0

(3)

The element in the first row and the first column is the probability of sending a bit of 0 and receiving it correctly.
The element in the first row and the second column is the probability of sending a bit of 0, but incorrectly
detecting it as 1.
November 25, 2013

DRAFT

A. Bit Error Probability for One-Branch Model

In CSRC for 16-QAM, each hop detects two bits. The transmission error probability for each bit is not always
the same, because they are detected at different nodes (the relay node detects the first two bits and the destination
detects the last two bits). The error probabilities for the bits are discussed separately and followed by the overall
bit error probability of CSRC in the following.
For the first two bits, their error probabilities denoted as P1 and P2 , they are detected at R, transmitted to D,
and finally demodulated by D. With the assumption that SNR of all channels are identical, the channel transition
matrixes of the channel between S and R and the channel between R and D are identical. Therefore, the transition
matrix for the transmission of the two bits by CSRC strategy, denoted as T12 , is
T12 = T T

P02 + (1 P0 )2
=
2P0 (1 P0 )

2P0 (1 P0 )
P02 + (1 P0 )2

Thereby, the error probability of the first two bits is

P1 = P2 = 2P0 (1 P0 )

(4)

For the last two bits whose error probabilities denoted as P3 and P4 , the detection depends on the previous
detection of the first two bits. To assure the correction of the detection, the first two bits must be received and
demodulated correctly at D in order to locate the region of the signal in the constellation map. In other words, the
detection of QPSK hint H must be right. With Formula (2), the transition matrix for the transmission of the last
two bits with CSRC strategy, denoted as T34 , is

1 Ps
T34 =
Ps

1 P0

1 Ps
P0
Ps

P0
1 P0

Define Pa = (1 Ps ) (1 P0 ) + Ps P0 , and Pb = (1 Ps ) P0 + Ps (1 P0 ). Then

T34

Pa
=
Pb

Pb
Pa

Thus, the error probability of the last two bits is

P3 = P4 = Pa = (1 Ps ) P0 + Ps (1 P0 )

(5)

The symbol error probability of CSRC strategy is

Psymbol = 1

4
Y

(1 Pi )

(6)

i=1

The bit error probability can be approximated by

Pbit Psymbol /log2 M

(7)

where M is the order of the modulation. In this example, since 16-QAM is used, M equals to 4.
November 25, 2013

DRAFT

10

Based on Formula (7), the bit error performance of CSRC strategy and conventional AF strategy in one-branch
16-QAM case is illustrated in Figure 9 against SNR on symmetric links with the same condition.CSRC shows clear
performance gain over the conventional AF relaying and combining strategy.

10
10

BER

10
10
10
10
10

Fig. 9.

CSRC
AF

-1

-2

-3

-4

-5

-6

5
SNR

10

Comparison of theoretical bit error probability of 16-QAM modulated transmission between CSRC and conventional amplified and

forward strategy in relay channel

B. Bit error probability for Two-Branch Cooperative Model

To analyze the bit error probability of CSRC in the collaborative network model, we take a simple collaborative
case of two branches with N = 2. We will analyze the performance at different values of N in the next section.
In the collaborative network model, the effect of combing at the destination must be considered. Suppose MRC is
used for combining, which maximizes the SNR with proper weights assigned to each branch. The SNR of the MRC
result equals to the sum of SNR of each branch. In our analysis, d denotes the SNR after MRC at the destination;
1 denotes the SNR of the upper branch; 2 denotes the SNR of the lower branch. The SNR after MRC at the
destination is d = 1 + 2 . The transition matrix of the relay-destination channel is impacted by d . To obtain the
transition matrix of the relay-destination channel, we first define
p 
P0 0 = Q
2d , and

0
0
1

P
P
0
0

T0 =
1 P0 0
P0 0
The transition matrix for the transmission of the first two bits, denoted as T12 0 , is
T12 0 = T T 0
The transition matrix for the transmission of the last two bits, denoted as T34 0 , is

P
P
s
s
T0
T34 0 =
Ps
1 Ps
Using these two transition matrices in the same way as what is formulated in Formulas (4), (5), (6) and (7) for
the one-branch case, we can obtain the bit error probability for the two-branch case. The bit error performance of
November 25, 2013

DRAFT

11

CSRC strategy and the conventional AF strategy in the two-branch case is illustrated in Figure 10 against SNR.
Refer back to the one-branch performance in Figure 9, CSRC improves even more over the conventional AF in the
collaborative model. As expected, both of them gain performance improvement from collaborative communications.

10

AF
CSRC
10

-2

BER

10

-1

10

10

10

Fig. 10.

-3

-4

-5

5
SNR

10

Comparison of theoretical bit error probability of 16-QAM modulated transmission between CSRC and conventional amplified and

forward strategy in coorperative channel

C. Bit error probability for N -Branch Cooperative Model

It would be of utmost interest to see the effect of the order of diversity, namely the number of relay branches, on
CSRC performance. The destination uses MRC to combine the multiple copies of both the signal and hint coming
Pn
through N branches. The SNR after MRC at the destination is n = i=1 i . With this SNR, the bit error probability
can be deduced by the same way as the deducing of two branches cases. The results are shown in Figure 11 and show
that CSRC outperforms AF in all collaborative cases. An interesting and important observation is that both CSRC and
AF benefit in performance as the order of diversity increases. However, the collaborative gains end when they reach
their maximum (the case in our analysis occurs at the order of diversity of 3). In other words, collaborative diversity
does improve the performance, but after the gain is fully exploited, more collaborative diversity only wastes the
power consumption. This observation hints that when we consider collaborative communication, we should consider
the gain capacity, rather than blindly involving as many collaborative links as possible. The gain capacity is highly
constrained by the channel condition of source-relay links. As shown in Figure 11, the performance converges to
the limitation that indicates the BER performance of the source-relay links.
VI. P ERFORMANCE E VALUATION
We have conducted extensive evaluations on the performance of the proposed CSRC with both simulation and an
implemented Software Defined Radio (SDR) testbed. The performance of CSRC is compared with the conventional
November 25, 2013

DRAFT

12

10

BER

10

10

10

-1.1

AF
CSRC
-1.2

-1.3

-1.4

Fig. 11.

3
Diversity

Comparison of theoretical bit error probability of 16-QAM modulated transmission between CSRC and conventional amplified and

forward strategy in different diversity setting.

cooperative schemes of amplified-and-forward relaying [6], because both CSRC and the conventional methods do
not have the error control coding so that the effect of coding gain can be eliminated for fair comparisons.
A. Simulation Evaluation
Numerical results have been obtained with Monte Carlo simulations. The performance is measured in bit error
rate (BER) upon SNR. The simulations are based on AWGN channels in different scenarios. The performance with
AWGN channels can be considered as the upper bound. We consider bot models shown in Figure 1: the one-branch
model consists of three nodes: one source, one relay and one destination and a two-branch (N =2) model consists
of four nodes: one source, two relay nodes and one destination. We select a two-branch model as the collaborative
case because in our settings, as shown in Figure 11, the diversity gain fast converges to the maximum after N =2.
We have simulated CSRC for two modulations: 16-QAM and 64-QAM. Four scenarios are evaluated as follows.
1) Symmetric One-Branch Channel: In the first scenario, three-node model, the upper case in Figure 1, is used,
with the same SNRs on two links, which vary from 0dB to 10dB. In this scenario, the channel conditions of two
links are symmetric. The performance is measured from poor to good channel conditions. Figure 12 shows the
BER performance of the proposed CSRC and the conventional amplified-and-forward (AF) cooperative strategy
in this scenario, where x-axis refers to the SNR on each link. The simulation result shows the strengths of the
proposed strategy in improving BER. In Figure 12, we observe that CSRC outperforms the conventional AF strategy.
Specifically, it improves the BER around 2 to 3 dB over the conventional AF strategy. The performance gap
between two strategies for 64-QAM is larger than for 16-QAM, which indicates that the performance gain of CSRC
becomes greater when higher order modulation is employed to achieve higher data rate. This is because higher
order modulation leads to larger errors and gives more space for CSRC to improve.
2) Asymmetric One-Branch Channel: In the second scenario, the SNR of the link between the destination and the
relay varies from 0 dB to 10 dB, but the SNR of the other link is kept constant as low (5 dB is set in experiments).
In this scenario, the channel conditions of two links are asymmetric. This scenario particularly evaluates the effect
of low SNR source-relay condition on the performance. For this scenario, Figure 13 compares the performance of
November 25, 2013

DRAFT

13

Fig. 12.

The BER comparison of CSRC and AF in the first scenario.

BER. The x-axis represents the SNR of the link between the relay R and the destination D. Unreliable relay links
can cause error propagation and highly degrade the performance of cooperative communication. CSRC outperforms
the conventional strategy when the SNRs between the relay node and the destination are high. That is because CSRC
decomposes the high order demodulation into binary demodulation under the cooperation of the relay nodes and
avoids the decoding errors that significantly degrade the performance of conventional schemes. Especially, for 64QAM, CSRC uses lower order demodulation, QPSK, to detect the signal received from the poor SNR source-relay
link and higher order demodulation, 16-QAM, to detect the signal received from the high SNR relay-destination
link. This adaptation greatly improves the cooperative performance under poor relay link circumstances.

Fig. 13.

The BER comparison of CSRC and AF in the second scenario.

3) Symmetric Two-Branch Channel: In the third scenario, four-node channel model, the lower case in Figure 1,
is considered and all the links in the upper and lower branches have the same SNRs, which range from 0 dB to 10
dB. In this scenario, the channel conditions of two branches are symmetric. In this scenario, both CSRC and AF
employ MRC to combine the signals from upper and lower branches. This scenario reveals the performance under
diversity that is not available in one-branch cases. The result is demonstrated in Figure 14. It can be observed that

November 25, 2013

DRAFT

14

CSRC greatly outperforms the conventional AF, especially when the order of modulation increases. The reason is
that the low order demodulation, QPSK, is employed at the relay node to ensure that the destination can safely
locate the symbol correctly. Although a high order demodulation, 16-QAM, is used to detect the rest part of data,
using MRC at the destination provides diversity gain to compensate the high order demodulation loss.

Fig. 14.

The BER comparison of CSRC and AF in the third scenario.

4) Asymmetric Two-Branch Channel: The fourth scenario is designed for asymmetric links in two branches: all
the links of the upper branch have relatively low SNRs (5 dB), but the SNRs of the links in the lower branch vary
from 0 dB to 10 dB. In this scenario, the effect of one low SNR branch on the performance is revealed. As in the
third scenario, both CSRC and AF employ MRC at the destination to exploit the diversity. The result is shown in
Figure 15. Although the low SNR source-relay channels degrade the performance of both CSRC and AF, it can be
observed that CSRC outperforms AF, because the relay nodes use QPSK to demodulate the signals received from
the poor SNR source-relay channels.

Fig. 15.

The BER comparison of CSRC and AF in the fourth scenario.

B. SDR Experiments
We have also evaluated the performance on a small SDR experiment is presented. The purpose of the experiments
on the SDR tested is to validate the feasibility of hardware implementation and to test the performance in real
November 25, 2013

DRAFT

15

wireless environments. The SDR tested consists of three USRP 210N nodes deployed as the one-branch case shown
in Figure 1 in an indoor laboratory environment. The distance of each hop between nodes is set as 5 meters.
We implemented both CSRC and AF [6] into the SDR testbed with 16-QAM tested. The experiments have been
conducted at 5GHz frequency band. To test the performance in different channel conditions, the antenna gains are
varied in the tests.
Two scenarios are considered. In one scenario, the hops are configured as symmetric links with three antenna
gain values, specifically 10 dB, 12.5 dB and 15 dB corresponding to poor, average and good channel conditions.
The results obtained from the SDR tested is illustrated in Figure 16. From the results, although in low antenna gain
the performance gain is marginal, CSRC demonstrates better BER performance than AF in the average antenna
gain setting and high antenna gain setting. In the second scenario that is with asymmetric links, the antenna gain
for source-relay transmission is set as 10dB and the antenna gain for relay-source transmission is varied among
10 dB, 12.5 dB and 15 dB. This scenario is used to observed the performance under poor source-relay channel
conditions. As depicted in Figure17, the performance gain between CSRC and AF becomes more visible as the
relay-destination channel becomes better. When both links are poor, CSRC does not help much, but when the
relay-destination becomes good, CSRC helps improve the performance by mitigating the suffer on the source-relay
hop.

0
-0.05
-0.10

CSRC
AF

-0.15
-0.22
-0.30

BER (log)

-0.40

-0.52

-0.70

-1.00

10

12.5

15
SNR

Fig. 16.

The BER comparison of CSRC and AF in SDR experiment.

VII. C ONCLUSION
This paper proposes a novel Cooperative Step-Wise Relaying and Combining strategy (CSRC) for cooperative
relay in VANETs. CSRC supports high order modulation to achieve high bit rate transmission while maintaining
low bit error rate performance with low order demodulation detections at cooperative relay nodes. The strengths
of CSRC are confirmed with theoretical analysis of bit error probability, extensive simulations and SDR testbed

November 25, 2013

DRAFT

16

0
CSRC
AF

-0.05

-0.10

BER (log)

-0.15

-0.22

-0.30

-0.40

Fig. 17.

10

12.5
SNR (dB)

15

The BER comparison of CSRC and AF in SDR experiment.

experiments. CSRC can be further extended to support other higher-order modulations such as 256-QAM over
multi-hop wireless networks.
VIII. ACKNOWLEDGEMENT
The authors would thank the support from the US National Science Foundation through the grants #1041292 and
#1041095 to perform this work.
R EFERENCES
[1] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.

Cambridge.

[2] A. Sendonaris, E. Erkip, and B. Aazhang, User cooperation diversity, part I: System description, IEEE Transaction on Communications,
vol. 51, no. 1, p. 19271938, 2003.
[3] , User cooperation diversity, part II: Implementation aspects and performance analysis, IEEE Transaction On Communications,
vol. 51, no. 1, p. 19391948, 2003.
[4] G. Kramer, M. Gastpar and P. Gupta, Cooperative strategies and capacity the independent and identically distributed theorems for relay
networks, IEEE Transaction on Information Theory, vol. 51, pp. 30373063, 2005.
[5] T. Tran, Y. Shin and O. Shin, Overview of enabling technologies for 3gpp lte-advanced, EURASIP Journal on Wireless Communications
and Networking, 2012:54.
[6] J. Laneman, D. Tse and G. Wornell, Cooperative diversity in wireless networks: Efficient protocols and outage behaviour, IEEE
Transactions on Information Theory, vol. 50, no. 12, pp. 30623080, 2004.
[7] D. Chen and J. N. Laneman, Modulation and demodulation for cooperative diversity in wireless systems, IEEE Transactions on Wireless
Communications, 2005.
[8] A. Stefanov and E. Erkip, Cooperative coding for wireless networks, IEEE Transaction On Communications, vol. 52, no. 9, 2004.
[9] H. H. Sneessens and L. Vandendorpe, Soft decode and forward improves cooperative communications, 3G 20056th IEE International
Conference on 3G and Beyond, 2005.
[10] K. Azarian, H. El Gamal, and P. Schniter, On the achievable diversity-multiplexing tradeoff in half-duplex cooperative channels, IEEE
Transactions on Information Theory, vol. 51, no. 12, p. 552572, 2005.

November 25, 2013

DRAFT

17

[11] Y. Li, B. Vucetic, T.F. Wong, and M. Dohler, Distributed turbo coding with soft information relaying in multihop relay networks, IEEE
Journal on Selected Areas in Communications, vol. 24, no. 11, p. 2040 2050, 2006.
[12] R. Thobaben, On distributed codes with noisy relays, In Proceedings of Asilomar Conference on Signals, Systems and Computers,, pp.
2629, 2008.
[13] A. Chakrabarti, A. de Baynast, A. Sabharwal and B. Aazhang, Low density parity check codes for the relay channel, IEEE Journal on
Selected Areas in Communications, vol. 25, no. 2, pp. 280291, 2007.
[14] J. Hu and T. M. Duman, Low density parity check codes for the relay channel, IEEE Transactions on Communications, vol. 6, no. 9,
p. 33843394, 2007.
[15] K. Gomadam and S. A. Jafar, Optimal relay functionality for snr maximization in memoryless relay networks, IEEE Journal on Selected
Areas in Communications, vol. 25, no. 2, pp. 390401, 2007.
[16] P. Weitkemper, D. Wbben, V. Khn and K. D. Kammeyer, Soft information relaying for wireless networks with error-prone source-relay
link, Proceedings of 7th International ITG Conference on Source and Channel Coding, 2008.
[17] J. Adeane, M. R. D. Rodrigues, and I. J. Wassell, Characterisation of the performance of cooperative networks in ricean fading channels,
12th Int. Conf. Telecommun., Cape Town, South Africa, 2005.
[18] T. Wang, A. Cano, G. Giannakis and N. Laneman, High performance cooperative demodulation with decode-and-forward relays, IEEE
Transactions on Wireless Communications, vol. 55, no. 7, pp. 14271438, 2007.
[19] B. Rankov and A. Wittneben, Spectral efficient protocols for half-duplex fading relay channels, IEEE J. Select. Areas Commun., vol. 25,
pp. 379389, 2007.
[20] P. Popovski and H. Yomo, Wireless network coding by amplify-and-forward for bi-directional traffic flows, IEEE Commun. Lett., vol. 11,
pp. 1618, 2007.
[21] R. Ahlswede, N. Cai, S.-Y. R. Li, and R. W. Yeung, Network information flow, IEEE Transactions on Information Theory, vol. 46, no. 4,
pp. 12041216, July 2000.
[22] M.C. Thomas and A.T. Joy, Elements of Information Theory (second edition).

November 25, 2013

WILEY.

DRAFT